朱宇航 史云鵬
摘 要: 在用冪法求取矩陣的特征值時(shí),將已知向量視為特征向量的線性組合,用矩陣對(duì)已知向量做左累乘的迭代運(yùn)算,相鄰迭代運(yùn)算向量分量之商就是特征值中絕對(duì)值最大者的近似值。通過對(duì)迭代過程的無窮小分析可知,在忽略二階無窮小時(shí),近似值與精確值之差呈現(xiàn)等比數(shù)列規(guī)律,依次進(jìn)而設(shè)計(jì)了加速算法,但其結(jié)果受二階及更高階無窮小的影響,其與精確值的差值在迭代運(yùn)算中仍呈現(xiàn)等比數(shù)列規(guī)律,再次應(yīng)用前面的加速算法,明顯提高了計(jì)算精度,達(dá)到了計(jì)入二階無窮小的效果,而計(jì)入二階無窮小的方程組是無解析解的。
關(guān)鍵詞: 二階矩陣 特征值 特征向量 加速算法
中圖分類號(hào): O151.21文獻(xiàn)標(biāo)識(shí)碼: A文章編號(hào): 1679-3567(2024)05-0073-03
A Method for Obtaining the Eigenvectors of Second-Order Matrix
ZHU Yuhang1 SHI Yunpeng2
1.China Machinery Technology (Beijing) Vehicle Testing Engineering Research Institute of China, Beijing, 102100 China; 2.China Productivity Center for Machinery Co., Ltd., Beijing, 100044 China
Abstract: When using a power method to obtain the eigenvalues of a matrix, a known vector is treated as the linear combination of eigenvectors, the matrix is used to perform an iterative operation of the left multiplication of the known vector, and the quotient of the vector components of adjacent iterative operations is the approximate value of the one with the largest absolute value in the eigenvalues. Through the analysis of infinitesimals in the iteration process, it can be seen that when ignoring second-order infinitesimals, the difference between the approximate value and the exact value presents a geometric progression rule. In turn, an acceleration algorithm is designed, but its result is affected by second-order and higher-order infinitesimals, and the difference between it and the exact value still presents a geometric progression rule in the iteration operation. The previous acceleration algorithm is applied again, the calculation accuracy is significantly improved, achieving the effect of including the second-order infinitesimals, and the system of equations included in the second-order infinitesimals has no analytical solution.
Key Words: Second-order matrix; Eigenvalue; Eigenvector; Accelerated algorithm
動(dòng)平衡機(jī)中需要精確求取具有正弦波的一次諧波系數(shù)時(shí),需要對(duì)脈沖干擾信號(hào)做濾波處理,傳統(tǒng)的低通濾波方法不能在數(shù)學(xué)上保證求取的一次諧波的精度。因此提出了新的方法:用擬合的數(shù)據(jù)替代干擾區(qū)間的數(shù)據(jù)進(jìn)而求取新的一次諧波系數(shù),逐次迭代計(jì)算,直至相鄰兩次的迭代值充分接近為止。本文涉及的迭代計(jì)算其本質(zhì)就是用冪法求矩陣的特征值和特征向量。本文只討論比較簡單的二階矩陣。其方法亦可推廣到高階矩陣。
特征方程的次數(shù)就是矩陣的階數(shù),三階矩陣的特征方程是一個(gè)三次代數(shù)方程,一般采用卡旦解法[2],對(duì)應(yīng)于四階矩陣的四次方程則用費(fèi)拉里法求解,而五次及以上的方程無一般解法,所以五階及更高階的矩陣的特征值只能用數(shù)值解法求出近似值而無解析解。
這就是二次加速算法,是對(duì)加速算法忽略二階無窮小的近似補(bǔ)償。
本文中利用特征值和特征向量性質(zhì)的方法求特征向量,有時(shí)能比解齊次線性方程組非零解的方法簡單些,并且本方法比較直觀,比較適合于編軟件。對(duì)于本文算例,提出的算法的收斂速度略好于Aitken和威爾金森算法,但比這兩種算法簡便,在采用冪法求矩陣的特征值時(shí),可以考慮采用本文提出的二次加速算法。
參考文獻(xiàn)
[1]孟憲萌,???利用特征矩陣求實(shí)對(duì)稱矩陣的特征向量[J].高等數(shù)學(xué)研究,2021,24(1):21-23.
[2]張仙鳳.淺談抽象矩陣特征值、特征向量的求法及其應(yīng)用[J].景德鎮(zhèn)學(xué)院學(xué)報(bào),2021,36(3):96-99.
[3]盧靖程.數(shù)值線性代數(shù)算法在工程軟件中的應(yīng)用與意義[J].科學(xué)與信息化,2023(3):61-64,69.
[4]雍龍泉.矩陣特征值與特征向量的幾何意義[J].陜西理工大學(xué)學(xué)報(bào):自然科學(xué)版,2021,37(5):80-85.
[5]馬麗娜,劉爍.淺談線性代數(shù)課程教學(xué)設(shè)計(jì):以“特征值與特征向量”為例[J].高等數(shù)學(xué)研究,2023,26(1): 95-97.